Google has made its SynthID Detector tool available to the public on October 7, 2026, through the website synthid.com. This tool checks whether an image, video, or audio file contains a SynthID watermark, which is added by Google and its partners to content created by artificial intelligence. Previously, only journalists, researchers, and media professionals could access the tool after signing up for a waiting list. The announcement was made by Pushmeet Kohli, vice president of Google DeepMind, the company's AI research division. To use the tool, users need to connect with a Google, Apple, or OpenAI account and upload the file. The site then checks for the presence of the SynthID watermark. The tool is available globally but only in English for now. SynthID was introduced by Google DeepMind in 2023. This technology embeds an invisible watermark into AI-generated content at the moment it is created. The watermark is not visible to the human eye or ear, but it can be detected using the SynthID Detector. According to Google, the watermark remains detectable even after the file is shared or modified in certain ways. This makes it more reliable than simple metadata, which can be easily removed. However, Google admits that the watermark can be bypassed, especially when it comes to text-based content. Since 2023, Google has expanded the use of SynthID to text generated by AI, then to images modified by AI in Google Photos. According to the company, more than 180 billion images and videos now carry this watermark, along with the equivalent of 240,000 years of audio content. In May 2025, Google reported slightly more than 10 billion marked contents across all formats. A year later, at the Google I/O 2026 event, it claimed more than 100 billion images and videos and 60,000 years of audio were marked. In May 2026, Google announced its plan to make SynthID the standard for detecting AI-generated content. Gemini, Google Research, and Chrome can already determine if an image contains the SynthID watermark, with the long-term goal of having a single watermark for all services. On October 7, 2026, Google launched the SynthID Detector website, which had been presented the previous year but was previously only available to professionals and media organizations. SynthID Detector works by allowing users to upload an image, video, or audio file. The tool then searches for the presence of the SynthID watermark, an invisible signature embedded in the content at the time of its creation. According to Google, the watermark remains intact even after cropping, color changes, or compression, making it impossible to remove. In addition to content generated by Google's Nano Banana, SynthID can also detect creations from OpenAI (including ChatGPT Images 2.0), NVIDIA, and Kakao. Google mentions that Apple will soon join the list of partners, though it has not provided a specific date. Several of Google's competitors have agreed to adopt SynthID as a standard. However, the SynthID Detector does not "recognize" AI content directly; it only detects the watermark. Other platforms, such as Midjourney, Meta, or xAI, are not part of the partnership, and some lesser-known tools can also operate without using SynthID. This means that individuals who want to bypass the system can still do so. However, the most popular tools will leave a trace. The situation with Apple is particularly interesting. At the WWDC 2026 event in June, Apple announced that images generated by Image Playground and photos edited with Apple Intelligence would include a hidden SynthID watermark. However, iOS 27 has already been released since September 14, and images created with Image Playground already carry this watermark. It is unusual that Google has not yet listed Apple as an official partner. As it stands, content created by tools that do not use SynthID will pass the test of the new tool without triggering an alert, even if it is entirely fake. If an image seems suspicious, users are encouraged to test it. A detected watermark is a strong indication that the content comes from an AI. However, a negative result does not prove the content is authentic.